Microbiology (1999), 145, 2625–2634
REVIEW
ARTICLE
Printed in Great Britain
Cloning and assembly strategies in microbial
genome projects
Lionel Frangeul,1 Karen E. Nelson,2 Carmen Buchrieser,1
Antoine Danchin,3 Philippe Glaser1 and Frank Kunst1
Author for correspondence : Frank Kunst. Tel : j33 1 45 68 88 69. Fax : j33 1 45 68 87 86.
e-mail : fkunst!pasteur.fr
Laboratoire de Ge! nomique des Microorganismes Pathoge' nes, Institut Pasteur, 25 rue du Dr Roux,
75724 Paris Cedex 15, France
1
2
The Institute for Genomic Research, 9712 Medical Center Drive, Rockville, MD 20850, USA
Unite! de Re! gulation de l’Expression Ge! ne! tique, Institut Pasteur, 28 rue du Dr Roux,
75724 Paris Cedex 15, France
3
Keywords : genome sequencing, microbial genomes, cloning strategies, shotgun,
bioinformatics
Overview
The knowledge of an entire genome sequence not only
provides a wealth of data, but also specific information
that can not be obtained by other approaches. Only after
the completion of genome projects did it become obvious
that many genes had not been identified by classical
genetics. These included many genes that do not share
similarities with known genes in databases, paralogous
genes leading to redundancy of gene function, genes
corresponding to potential new drug targets (Arigoni et
al., 1998) and inactive genes resulting from reductive
evolution (Andersson et al., 1998). Whole-genome
sequencing has revealed important information on the
organization of genomes. These include the GjC
content and its variation within the genome, leading to
the identification of DNA regions acquired by horizontal
gene transfer, the presence of repeated elements or
insertion elements, the discovery of new pathogenicity
islands in pathogenic bacteria and the identification of
new operons, genome polarity, and the identification of
origin(s) of replication. Genome projects also can be
starting points for other projects such as metabolic
reconstruction (Selkov et al., 1997) and the systematic
functional analysis of all of the genes of an organism
which are, for example, currently being carried out for
Saccharomyces cerevisiae (Dujon, 1998) and Bacillus
subtilis (see the INRA web site ; Table 1).
Several projects involving genome sequencing of model
organisms, for example S. cerevisiae (Goffeau et al.,
1997), Escherichia coli (Blattner et al., 1997) and B.
subtilis (Kunst et al., 1997), have been completed using
a large amount of previously available information,
including the genetic and physical maps of the organisms
as well as ordered large-insert libraries of cosmid,
lambda, YAC (yeast artificial chromosome) or BAC
(bacterial artificial chromosome) clones.
Nowadays, substantial preliminary data are no longer
necessary for the initiation of a genome project. However, a prerequisite for a successful project is detailed
pre-planning, and some basic knowledge of the bacterium at the beginning of the project is helpful. Features
such as genome size, GjC content, the presence of a
circular or linear chromosome and the eventual presence
of (circular or linear) plasmids should be taken into
account. The strategies to be followed should be very
carefully considered, since midstream changes are time
consuming and costly. In particular, the choice of the
cloning strategies is crucial, since the construction of
gene banks of poor quality may be unnoticed in the
beginning of the project and lead, at a later stage, to
assembly difficulties (see below).
The aim of a microbial genome project is to construct,
from 500–800 bp sequence reads containing about 1 %
mistakes, a genome sequence of several megabases with
an error rate lower than 1 per 10 000 nucleotides. Because
of technical improvements, including the availability of
new software, automated sequencers and low computation costs, a microbial genome project can be
completed in a small laboratory within two years.
In the first section of this review we will present a
summary of the various cloning and sequencing
strategies used in genome projects, including random
and ordered approaches. In the following sections we
will focus on the whole-genome shotgun-sequencing
approach.
0002-3411 # 1999 SGM
2625
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Mon, 19 Jun 2017 05:13:41
L. FRANGUEL and OTHERS
Table 1. Web sites cited in the text
Organization
Web address (URL)
The Institute for Genomic Research
(TIGR)
GOLD, University of Illinois*
http :\\www.tigr.org\
http :\\www.ebi.ac.uk\research\cgg.genomes.html
http :\\geta.life.uiuc.edu\"nikos\genomes.html
http :\\www.invitrogen.com\
http :\\www.tree.caltech.edu\
http :\\www.pasteur.fr\recherche\unites\gmp\
Invitrogen
Caltech Genome Research Laboratory
Laboratory of Genomics of Microbial
Pathogens (GMP)
Stratagene
University of Oklahoma†
Clontech Laboratories
Phrap, University of Washington
INRA
http :\\www.stratagene.com\
http :\\www.genome.ou.edu\bigIdyesIbact.html
http :\\www.clontech.com\
http :\\bozeman.mbt.washington.edu\
http :\\locus.jouy.inra.fr
* GOLD, Genome On Line Database.
† Sequencing of bacterial genomic DNA with ABI’s big dye terminators.
Genome-sequencing strategies
Two genome-sequencing strategies have frequently been
used (Fig. 1). The first is the ordered-clone approach
that uses a large-insert library to construct a map of
overlapping clones covering the whole genome ; selected
clones are then sequenced to obtain the whole-genome
sequence. The second strategy is direct shotgun
sequencing, which does not require preliminary data
(such as a map) before the sequencing phase.
Ordered-clone approach
Several methods are used to order clones, including
restriction fingerprinting and hybridization mapping.
Fingerprinting is a procedure for clone comparison
based on the matching of characteristic restriction
fragment sets (‘ fingerprints ’). If the fingerprints of two
clones are found to share a significant number of
restriction fragments, it may be assumed that these two
clones overlap and form a contiguous sequenced region,
called a contig (Gregory et al., 1997). This method has
been successfully applied during the Caenorhabditis
elegans (Ellis et al., 1986) and Mycobacterium tuberculosis projects (Cole et al., 1998). Bioinformatic tools
such as ,  and  (Gillett et al., 1996 ;
Soderlund et al., 1997) may be of help for the construction and editing of clone maps and contigs, using
fragment size data from multiple complete endonuclease
digestions of a set of clones.
Hybridization is a simple and rapid method for
identifying stretches of homologous DNA, and a large
number of clones can be analysed and ordered concurrently. With this method, an ordered cosmid library
covering nearly the whole genome (8 Mb) of Streptomyces coelicolor has been constructed (Redenbach et
al., 1996). Chromosomal DNA was digested with a rarecutting endonuclease, and the resulting gel-purified
DNA fragments used as probes to order the cosmid
library. Cosmid sublibraries corresponding to each
DNA fragment were thus defined and the clones aligned
with the help of labelled riboprobes corresponding to
the insert ends. These probes can be easily prepared
since the cloning site of the vector is flanked by T3 and
T7 promoters adjacent to the insert ends. Management
of the data may be facilitated by using software such as
Cloneplacer (Singh & Krawetz, 1995).
Random-sequencing approach
Currently, the most widely used strategy for the
sequencing of a microbial genome is that of wholegenome shotgun sequencing. A large number of clones,
from libraries representative of the whole genome, are
sequenced and assembled into contigs. The contigs are
then joined together using a variety of methods to obtain
the whole genome sequence in a single contig. It should
be kept in mind that 90–95 % of the sequence, collected
during the shotgun phase of the project, is usually
assembled into several hundred contigs without any
difficulty. However, the obtention of the final 5–10 % of
the sequence and its assembly into a single contig is an
important task that requires careful planning, as is
outlined throughout this review.
With the use of the random-sequencing approach, the
complete genome sequences of several organisms have
been obtained, including Haemophilus influenzae
(1n83 Mb), Helicobacter pylori (1n66 Mb), Archaeoglobus fulgidus (2n18 Mb) and Thermotoga maritima
(1n86 Mb) (Fleischmann et al., 1995 ; Klenk et al.,
1997 ; Nelson et al., 1999 ; Tomb et al., 1997 ; see also the
TIGR and University of Illinois web sites ; Table 1). This
method has now been adopted by many laboratories for
the completion of genomes in the size range of 2–6 Mb,
and possibly also beyond this size since the utilization of
this method has been discussed for the human genome
(Green, 1997 ; Weber & Myers, 1997).
In the following sections of this review we describe the
various phases of a whole-genome shotgun-sequencing
2626
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Mon, 19 Jun 2017 05:13:41
Microbial genome sequencing
Bacterial
chromosome
Smallinsert
library
Direct
wholegenome
shotgun
strategy
Large-insert library
Mapping
Sequences
of smallinsert
library
Shotgun sequencing
or partial sequencing
of each selected
large insert
Assembly
.....................................................................................................
Wholegenome
sequence
Closure
strategies
(according to
the map)
project, including the cloning step, the randomsequencing phase, the assembly of sequences into contigs
and the closure phase where the contigs are assembled
into a single contig spanning the entire genome sequence.
The precautions to be taken to ensure completion of a
genome-sequencing project and the difficulties encountered during each phase of the project will be
discussed.
Library construction
Experience from completed genome projects has demonstated the need for several libraries (Fleischmann et
al., 1995). The recommended procedure is to construct a
small- and a large-insert library concurrently. The smallinsert library is used for an appropriate coverage of the
genome and the large-insert library to obtain a ‘ scaffold ’
of the genome which is used during the closure phase.
The first precaution to be taken before starting the
construction of a library is to ensure that the genome of
the chosen organism is stable under the culture conditions used. In bacteria with a short doubling time in
rich media (less than 40 min), for example E. coli, B.
subtilis and many others, the region surrounding the
origin of replication will be overrepresented four to
eight times compared to the replication termination
region. To avoid such a potential bias in gene representation in genomic DNA libraries, it may be
advisable to extract genomic DNA from a culture in the
Sequences
of contigs
Fig. 1. Strategies used to obtain wholegenome sequences : the ordered-clone
approach with the help of a map
constructed from a large-insert library and
the direct shotgun approach with the
utilization of a small-insert library.
early stationary phase. However, an argument against
this approach is that the cell walls of some stationaryphase bacteria are more difficult to lyse than those of
exponentially growing bacteria. Additionally, cloning
problems may result from the base composition (AjTrich regions) or structural features of the DNA (palindromic sequences). Such problems may be solved by
generating libraries with very small inserts (200–500 bp)
(McMurray et al., 1998), or PCR amplification. Special
precautions are recommended for the manipulation and
cloning of AjT-rich DNA. These include extraction of
DNA under high-salt conditions to prevent DNA from
melting at a high temperature, the use of a reduced
extension temperature in PCR and the use of
transposon-insertion methods for the closure of gaps
that are very rich in AjT (Gardner et al., 1998).
Small-insert libraries
It is important to construct a small-insert (1–2 kb)
library with a single genomic DNA insert in each
recombinant clone. The presence of multiple insertions
in a given clone would cause genome-assembly artifacts.
To avoid tandem inserts of two independently cloned
fragments (co-cloning), Fleischmann et al. (1995) developed a procedure based on two steps of size selection.
The BstXI non-palindromic cloning method is another
efficient strategy which minimizes co-cloning. This
2627
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Mon, 19 Jun 2017 05:13:41
L. FRANGUEL and OTHERS
method has been applied to several genome projects
(Smith et al., 1997) and involves the use of special
vectors containing a polylinker region with two BstXI
sites, for example the ColE1 derivatives pcDNA1.1\
Amp or pcDNA2.1 (Invitrogen ; Table 1). After BstXI
cleavage, the vector fragment containing non-palindromic TGTG-3h overhanging ends is purified and
ligated to chromosomal DNA that has been fragmented
by nebulization (Bodenteich et al., 1994). However,
prior to ligation, gel-purified fragments in the range of
1–2 kb have their overhanging ends filled-in by T4 DNA
polymerase, then they are ligated to CACA-3h adaptors
and repurified. Under these conditions, a high proportion of the obtained recombinant plasmids contain
single DNA inserts because the vector fragment can not
recircularize and the probability of adaptor addition to
a single blunt-ended DNA fragment is substantially
higher than the ligation of two blunt-ended DNA
fragments.
Large-insert libraries
High-level expression of genes whose products are toxic
to the E. coli host (Kurland & Dong, 1996) is a major
cause of cloning problems, especially with large-insert
libraries (McMurray et al., 1998). This has been shown
to be the case for genes encoding membrane proteins
(Beck & Bremer, 1980 ; Beucher & Sparling, 1995 ; Luo
et al., 1997). The problem of overexpression is particularly severe when genes from Gram-positive bacteria
are cloned in a Gram-negative host. Gram-positive
bacteria use ribosomal binding sequences which are
close to the consensus sequence for efficient translation
initiation (Farwell et al., 1992 ; Isono & Isono, 1976).
This can be avoided, or at least reduced, by using a lowcopy-number vector, in particular when a large-insert
library is required to provide a scaffold of the entire
bacterial genome.
Large fragments (20–300 kb) are usually cloned in
lambda, cosmid, fosmid or BAC vectors. The main
advantages of BAC vectors, which have the origin of
replication of the E. coli F factor (Shizuya et al., 1992),
are the potential to clone inserts of up to 300 kb and
their stable maintenance. Moreover, BAC vectors have a
copy number of one to two copies per cell, reducing the
potential for recombination between cloned DNA fragments and avoiding counterselection due to overexpression of genes (Brosch et al., 1998). Their drawback
is the somewhat laborious construction of BAC
libraries : shearing of DNA must be avoided and n vitro
manipulations such as restriction digestions are therefore performed in agarose plugs. A convenient BAC
vector (pBeloBAC11) has been described by Kim et al.
(1996) and detailed information on this vector, as well as
protocols for BAC library construction, are available
from the California Institute of Technology web site
(Table 1). Alternative low-copy-number vectors are
derivatives of pRK290, which exist at 5–8 copies per
chromosomal equivalent in E. coli. Such vectors have
been shown to maintain stably foreign DNA up to
300 kb (Tao & Zhang, 1998). Fosmid vectors are BAC
vectors that contain cos sites, facilitating their in vitro
packaging into lambda (Kim et al., 1992). YAC vectors
have also been used for the cloning and stable maintenance of bacterial genomic DNA (Azevedo et al.,
1993) ; however, the low yield of purified YAC recombinant DNA and contamination with yeast genomic
DNA have limited their utilization.
Other versatile low-copy-number plasmid vectors are
pSYX34 (Xu & Fomenkov, 1994) and pHSG575
(Takeshita et al., 1987). These vectors, which are
maintained at about 5 copies per cell, are derivatives of
pSC101 (Cohen et al., 1973), contain multiple cloning
sites and allow for lacZ α-complementation. We have
found pSYX34 to be an efficient vector for the construction of a large-insert library (6–12 kb) for the DNA
of Gram-positive bacteria. In the case of the Listeria
monocytogenes sequencing project, carried out by a
consortium of European laboratories (GMP web site ;
Table 1), the partial fill-in method was used for the
construction of a large-insert library (Korch, 1987). The
pSYX34 vector was cleaved with SalI and the ends
partially filled-in using Klenow polymerase to create 5hTC overhangs. Chromosomal DNA from L. monocytogenes was partially digested with Sau3AI and
partially filled-in to create 5h-GA overhangs. After in
vitro ligation and transformation, 86 % of the clones
displayed a white phenotype and about 90 % of these
contained inserts in the expected size range of 6–12 kb
(our unpublished results). In contrast, when using a
high-copy-number plasmid such as pBluescript (Stratagene ; Table 1), the percentage of white clones was much
lower (9 %), and only 25 % of these contained inserts.
These results show the advantage of a low-copy-number
vector such as pSYX34 for the construction of largeinsert libraries, especially in the case of DNA from
Gram-positive bacteria. Cloning difficulties associated
with high copy ColE1-type vectors can also be avoided
using E. coli strains in which the copy number of these
plasmids is lowered. The copy number of ColE1
derivatives is strongly reduced in the pcnB cya mutant
TP611 (Glaser et al., 1993) and is reduced about fourfold
and 10-fold in the ABLE C and K strains, respectively,
commercially supplied by Stratagene (Table 1).
Assembly phase
During the assembly and closure phases, both computational and experimental results are crucial for
obtaining the entire genome sequence. The various
software tools used during these steps are listed on the
GMP web site (Table 1). It should be kept in mind that
the choice of an appropriate bioinformatics package
should be made at the beginning of the project, since
changing to another package generally leads to a vast
amount of additional work.
The assembly phase is composed of three major steps :
the conversion of the data from automated sequencers
to nucleotide sequences, the utilization of these sequences in the assembly process and the continuous
assessment of this assembly process.
2628
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Mon, 19 Jun 2017 05:13:41
Microbial genome sequencing
Utilization of the data supplied by automated
sequencers
A large diversity of programs or packages has been
developed for the various stages from gel-image analysis
to assembly editing. The analysis of the gel image is
generally performed by the software supplied with the
sequencer. However, it can also be analysed by other
software tools such as Getlanes (Cooper et al., 1996) or
GelImager (Giddings et al., 1998), which allow an
automatic or semi-automatic separation of low-quality
sequences from those of adequate quality. The following
step, ‘ base-calling ’, is critical for obtaining high-quality
sequences suitable for subsequent annotation of the
contigs. In the past, for each chromatogram extracted
from the gel image, users had to discriminate between
high and low quality sequence areas. Recently, the
development of programs such as BaseFinder (Giddings
et al., 1998) or Phred (Ewing & Green, 1998 ; Ewing et
al., 1998) has led to significant improvements in these
procedures. These software tools analyse the traces of
each sequence lane automatically and deliver a nucleotide sequence with an accuracy value for each nucleotide.
During assembly, the accuracy value associated with
each base allows the use of the entire sequence extracted
from the traces. The weight of one base in the multialignment is proportional to the accuracy value provided
by the base-caller.
The appropriate functioning of bioinformatic tools may
depend on the type of automated sequencer used. Some
base-callers have been created and validated on the
chromatograms generated by one type of automated
sequencer, leading to a slight bias of the software to the
type of apparatus used. For instance, in our hands, the
Phred base-caller produces better results with PE-ABI
chromatograms compared to the standard chromatogram format originating from sequencers using the
single-colour fluorescent technology.
Before the assembly step, cloning vector sequences are
removed from the sequences supplied by the base-caller.
These vector sequences can be masked automatically by
several software tools such as Cross-match which uses
the Smith and Waterman algorithm (University of
Washington web site ; Table 1).
Assembly step
The assembly of more than 10 000 sequences into a few
contigs, typically required for a bacterial genome
project, is based on complex algorithms. The following
need to be taken into account : the error rate of each
sequence, the use of the generated sequence itself or its
complement, and the presence of repeated sequences
whose missassembly needs to be avoided. Computer
tools to solve these problems were first described by
Staden (1979, 1982) and have been intensively developed
over the last few years (Kececioglu & Myers, 1995). The
most recent algorithms perform pairwise comparisons
for all sequences, allowing automatic threshold selection
with respect to the decision of whether two sequences
overlap or not. Clusters of overlapping sequences are
constructed and consensus sequences are deduced from
these clusters. The advanced algorithms allow the
location of repeated and chimeric sequences. The most
widely used assembly software tools are the TIGRAssembler (Sutton et al., 1995), CAP2 (Huang, 1996),
GAP (Bonfield et al., 1995) and Phrap (P. Green,
University of Washington, Seattle, USA, unpublished).
To take advantage of each software, efforts have been
made to allow data exchanges between some of the
packages. For example, TIGRAssembler can create an
assembly-result file in Phrap format so that it can be
viewed with Consed (Gordon et al., 1998). Moreover,
tools such as Caftools (Dear et al., 1998) have been
created to facilitate such exchanges. In 1994, a comparison of 11 assembly software tools showed important
discrepancies between their results (Miller & Powell,
1994). However, to our knowledge, no recent comparisons have been performed on currently available
software tools.
Providing there is no cloning bias, the assembled DNA
fragments are located around the chromosome according to a Poisson distribution (Lander & Waterman,
1988 ; Fraser & Fleischmann, 1997). The fraction of the
genome which remains unsequenced on both strands, is
p le−nw/L
!
where n is the number of sequenced clones, w is the the
mean length of a sequence and L is the length of the
genome (in kb). A threefold coverage of the genome
(nw\Ll3) would therefore produce about 95 % (p l
0n05) of the entire genome sequence on the basis of! a
random distribution. This level of coverage should
already be sufficient for the discovery of genes of interest.
If the final goal of the project is the completion of a
genome sequence, a higher coverage, up to eightfold, is
needed for the accomplishment of the shotgun phase.
This would require about 15 000 reads of 500 bp per
megabase of genome length and would produce theoretically more than 99 % of the genome sequence. The
number of contigs, which equals the number of unsequenced regions (double-strand gaps), can be calculated as
no. contigslno. gapslne−nw/L (Fraser &
Fleischmann, 1997)
For a 4 Mb genome, the predicted number of contigs
after a sevenfold coverage is about 50 (nl56 000,
Ll4 000 kb, wl0n5 kb). In practice, the number of
contigs is generally higher since several criteria are not
taken into consideration in this calculation. The assembly software needs an overlap of several nucleotides
to link two clones and the representation of some DNA
regions may be biased due to cloning difficulties. In this
context, an important effort must be dedicated to the
assessment of the assembly, as explained below.
Assessment of the assembly
Several software tools such as Consed (Gordon et al.,
1998) are dedicated to editing the assembly results. They
show the structure of the assembly (the location of all
2629
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Mon, 19 Jun 2017 05:13:41
L. FRANGUEL and OTHERS
clones inside the contigs) with direct links to the traces
of each clone and allow tags to be added to the
sequences. Some of these tags change the accuracy value
attributed to the sequences of a clone, and consequently
affect the result of the next assembly run.
At low coverage levels of about onefold, intermediate
assemblies should be carried out as tests : a large number
of repeat sequences may occur in the genome and it is
preferable to find solutions to this problem at an early
stage of the project. The simplest way to avoid misalignment due to short repeated sequences is to obtain
sequence reads that span most of the repeats. After
assembly, most long repeats are found not to be exact
repetitions. In this case, it is important to check contig
regions containing several clones with mismatches in
high-quality (i.e. high confidence values) sequence areas.
If these regions correspond to a false assembly, the
assembly software can be re-run with instructions to
separate these clones.
Similarly, the locations of the end sequences of each
clone have to be verified. In most cases, these sequences
are part of the same contig containing the first sequence
and the complement of the second sequence. A further
requirement is that these two sequences are separated by
a distance compatible with the cloned-insert length.
Two sequences separated by an inconsistent distance
and\or oriented in the same sense indicate a potential
misalignment. A visual inspection of the corresponding
regions should be performed and the misalignment
corrected using tags in the sequences to allow the
software to perform a better assembly.
There are no clear bases for deciding when to shift from
the random strategy to the closure phase. This decision
depends on the following considerations : does the contig
number decrease sufficiently in proportion to the number of new sequences added to the assembly ; is the mean
coverage of each contig sufficient ; are the closure
strategies expected to be successful at an acceptable cost
or would it be more advantageous to delay the shift from
the random to the closure phase ?
A thorough analysis of the sequences obtained by the
assembly software allows for an effective choice of the
moment to start the closure phase, which of course can
be performed concurrently with the end of the shotgun
sequencing phase.
Closure phase
After contig assembly, the first step is to order the
contigs and to try to link them together with specific
PCR products or cloned inserts that span each gap. The
resulting contigs that are still unlinked are extended
using methods as described below. Finally, a single
contig is obtained. Fig. 2 summarizes the various
methods used to identify the potential neighbours of
contigs.
During the random phase, both ends of small- and largeinsert clones are sequenced. Thus, if the terminal
(a)
End of contig1 Small insert
Sequence with
forward primer
End of contig2
Sequence with
reverse primer
(b)
Contig1
Large insert
Sequence with
forward primer
Contig2
Sequence with
reverse primer
(c)
Other genome already sequenced
(d)
End of contig1
End of contig2
Rare-cutting
endonuclease site
Rare-cutting
endonuclease site
Large genomic DNA fragment
Contig1
Contig2
Contig3
.................................................................................................................................................
Fig. 2. Methods for the construction of supercontigs. (a)
Contigs sharing sequences with a linking small-insert clone. (b)
Contigs sharing the end sequences of a linking clone from a
large-insert library. (c) Contigs sharing the same operon (or
gene) in another entirely sequenced genome. (d) Contigs
identified by hybridization to be located on the same large
genomic fragment. The symbols used are : cloned insert of the
linking clone (rectangle with dotted lines) ; sequences
performed on these clones (arrows) ; known sequences (black
boxes) ; unknown sequences (white boxes) ; similarity detected
by hybridization (iiiiiii) ; similarity detected by BLAST
(///////).
sequences of a single clone belong to different contigs, it
is highly probable that these two contigs are neighbours
(Fig. 2a, b). However, the orientation of the sequences
and the distance to the end of the contig should be
compatible with the size range of the inserts. These
clones are called ‘ linking clones ’ and the resulting
contigs are ‘ supercontigs ’. The probability that one
clone can be used as a linking clone increases with its
size. Therefore large-insert libraries are more useful at
this stage in the assembly process.
Whatever the coverage of the genome, physical gaps will
remain due to unclonable regions. However, the contig
order can be deduced from the analysis of coding
sequences (CDSs) predicted for each contig. If the
comparison between protein databases and the CDSs
from contig ends show that the ends of two contigs
encode different parts of the same protein, it may be
assumed that these contigs are neighbours. In addition,
the gene order in certain operons may also be conserved
between the genome being sequenced and the sequenced
genome of a related organism, hereafter called a control
2630
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Mon, 19 Jun 2017 05:13:41
Microbial genome sequencing
genome (Fig. 2c). In this way, a control genome may be
of help in the completion of the genome under study. For
the informatic part of this strategy we have developed a
software tool, called . For this program, the 
database of the contig sequences created by Phrap is
used as the input file. The user must choose the length of
contig ends to be taken into account for the calculation
(e.g. 1000–2000 nt).  then translates the largest
ORF in this terminal region and searches for the
counterpart of this protein in the protein database of the
control genome with  (Altschul et al., 1997). If a
significantly similar protein is found (20 % identity
over the total length of the smaller of the two proteins),
 searches for the location of the gene encoding this
protein in the control genome. Finally, all pairs of
selected genes from the genome under study whose
counterparts are less than 10 000 nucleotides apart in the
control genome, are transferred to a results file. This file
lists all these pairs with their homologous contig ends,
which are therefore likely to be neighbours. In the case
of L. monocytogenes, the protein sequences were
systematically compared with those of B. subtilis, whose
genome has already been completed (Kunst et al., 1997).
At least 20 % of a sample of 220 in silico predicted contig
links were shown to be authentic, since these have been
confirmed experimentally by the identification of largeinsert clones linking these contigs (our unpublished
data). With the increasing number of fully sequenced
genomes, it is likely that this strategy will be applicable
to many other projects.
If computational approaches fail to predict links between the remaining contigs, experiments may be
required. Southern-hybridization experiments, using
probes corresponding to different contigs, may show
that some of these contigs are located on the same large
restriction fragment (Fig. 2d). Such fragments can be
obtained after digestion of chromosomal DNA with a
rare-cutting endonuclease and separated by PFGE
(Ramos-Diaz & Ramos, 1998 ; Umelo & Trust, 1998 ;
Ze-Ze et al., 1998). Other techniques for rapid mapping,
including optical mapping and HAPPY mapping may be
envisaged (Cai et al., 1998 ; Piper et al., 1998). These
experimental approaches are time-consuming and may
be unnecessary in the case of small genomes.
All potential neighbour predictions have to be verified
by standard or long-range PCR (Barnes, 1994). For the
contigs without identified neighbours, gaps in genomic
sequences may be bridged by chromosomal-walking
methods. Direct sequencing of the chromosomal DNA
template (Heiner et al., 1998 ; see also the University of
Oklahoma web site ; Table1) has proved to be a powerful
technique, allowing the rapid completion of the T.
maritima genome project (Nelson et al., 1999). Another
method for closing gaps is inverse PCR (standard and
long-range ; Offringa & van der Lee, 1995). This method
allows the extension of contigs by amplification of
regions flanking known DNA sequences. Both inverse
and long-range PCR have been used extensively for the
completion of the B. subtilis genome project (Bolotin et
al., 1996 ; Ogasawara et al., 1994 ; Petit et al., 1998 ; Rose
& Entian, 1996 ; Sorokin et al., 1996). An interesting and
efficient alternative to inverse PCR is ligation-mediated
PCR. This single-sided amplification method involves
the ligation of a linker to DNA restriction fragments and
subsequent amplification with a pair of primers, one of
which recognizes the linker and the other a genomic
sequence (Prod’hom et al., 1998). A refinement of the
adaptor-ligation method, allowing reproducible amplification without background problems, has been published by Siebert et al. (1995) and a GenomeWalker kit
based on this method has been commercialised by
Clontech (Table 1).
An extension of the PCR methods, combinatorial PCR
(multiplex PCR) has also been developed. It involves
PCR amplification using a mixture of (up to 32) primers
that are located in the vicinity of contig ends. These
primers are designed to obtain the simultaneous amplification of PCR products, each primer pair allowing
linkage of two contigs. Parallel PCR reactions are
carried out in which one primer at a time is left out from
the mixture. The absence of a given PCR product in
these control reactions will identify the primer left out as
being required for the synthesis of a particular product.
After identification of each pair of productive primers,
the contigs can be ordered and the gaps filled after
sequencing of the obtained PCR products (Sorokin et
al., 1996).
Whatever the PCR method used, software tools such as
Consed or more specific programs such as Primo (Li et
al., 1997) or Pride (Haas et al., 1998) automatically
choose the primers needed for the closure phase ; the
primers are chosen according to the usual criteria
(selectivity, melting temperature, etc.) and also according to the accuracy values of contig sequences to
ascertain that the primer sequences correspond to
genome sequences likely to be devoid of errors. Consed
can also display the regions that have to be covered by
additional sequences. These include regions with lowquality coverage, regions covered only by a single clone
and regions sequenced only by dye-primer technology in
a single orientation, since this technology leads to an
increased frequency of sequence compressions.
Conclusion
Genomics is a new field of biology which has expanded
very rapidly in the past few years. Technological
advances in automated sequencing and in molecular
biology (gene cloning techniques, construction of
genome maps, etc.) together with the development of
bioinformatics, has ensured the dynamism of this
domain of biology. About 20 genome projects have been
completed and more than 70 projects are in progress (see
the TIGR and the University of Illinois web sites ; Table
1). As more genome sequences become available, it is
obvious that comparative genomics will play an increasingly important role in the closure phase of a new
genome project. In this context, new software tools will
be developed to meet the demands of this rapidly moving
field of biology.
2631
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Mon, 19 Jun 2017 05:13:41
L. FRANGUEL and OTHERS
The completion of a genome sequence represents the
end of one phase in the journey of discovery, and
important tasks still lie ahead. These include both in
silico analyses (homology searches) and experimental
analyses for the deduction of unknown metabolic
pathways, the prediction of operon structures and the
identification of regulatory signals. DNA microarray
techniques and two-dimensional proteomic analysis will
also be used for monitoring gene activity under various
environmental conditions. Additionally, the knowledge
of a complete genome may identify gene displacements
and horizontal transfers, which may help to trace
evolutionary networks and to define common ancestors.
The complete genome sequence of an organism provides
a wealth of information that will provide the basis for a
new generation of fundamental and applied biological
studies.
using long accurate PCR and three yeast artificial chromosomes.
Microbiology 142, 3017–3020.
Bonfield, J. K., Smith, K. & Staden, R. (1995). A new DNA
sequence assembly program. Nucleic Acids Res 23, 4992–4999.
Acknowledgements
Cooper, M. L., Maffitt, D. R., Parsons, J. D., Hillier, L. & States,
D. J. (1996). Lane tracking software for four-color fluorescence-
We wish to thank Rachel Purcell for helpful discussions and
critical reading of the manuscript.
The funding of the L. monocytogenes genome project (contract BIO4-98-0036), provided by the European Commission
in the framework of the Biotechnology program, is gratefully
acknowledged.
References
Altschul, S. F., Madden, T. L., Schaffer, A. A., Zhang, J., Zhang, Z.,
Miller, W. & Lipman, D. J. (1997). Gapped  and - : a
new generation of protein database search programs. Nucleic
Acids Res 25, 3389–3402.
Andersson, S. G. E., Zomorodipour, A., Anderssen, J. O. & 7 other
authors (1998). The genome sequence of Rickettsia prowazekii
and the origin of mitochondria. Nature 396, 133–140.
Arigoni, F., Talabot, F., Peitsch, M. & 7 other authors (1998). A
genome-based approach for the identification of essential bacterial genes. Nat Biotechnol 16, 851–856.
Azevedo, V., Alvarez, E., Zumstein, E., Damiani, G., Sgaramella,
V., Ehrlich, S. D. & Serror, P. (1993). An ordered collection of
Bacillus subtilis DNA segments cloned in yeast artificial
chromosomes. Proc Natl Acad Sci USA 90, 6047–6051.
Barnes, W. M. (1994). PCR amplification of up to 35 kb DNA with
high fidelity and high yield from lambda bacteriophage templates.
Proc Natl Acad Sci USA 91, 2216–2220.
Beck, E. & Bremer, E. (1980). Nucleotide sequence of the gene
ompA coding the outer membrane protein II of Escherichia coli
K-12. Nucleic Acids Res 8, 3011–3027.
Beucher, M. & Sparling, P. F. (1995). Cloning, sequencing, and
characterization of the gene encoding FrpB, a major ironregulated, outer membrane protein of Neisseria gonorrhoeae. J
Bacteriol 177, 2041–2049.
Brosch, R., Gordon, S. V., Billault, A., Garnier, T., Eiglmeier, K.,
Soravito, C., Barrell, B. G. & Cole, S. T. (1998). Use of a
Mycobacterium tuberculosis H37Rv bacterial artificial chromosome library for genome mapping, sequencing, and comparative genomics. Infect Immun 66, 2221–2229.
Cai, W., Jing, J., Irvin, B. & 8 other authors (1998). High-resolution
restriction maps of bacterial artificial chromosomes constructed
by optical mapping. Proc Natl Acad Sci USA 95, 3390–3395.
Cohen, S. N., Chang, A. C., Boyer, H. W. & Helling, R. B. (1973).
Construction of biologically functional bacterial plasmids in
vitro. Proc Natl Acad Sci USA 70, 3240–3244.
Cole, S. T., Brosch R., Parkill J. & 39 other authors (1998).
Deciphering the biology of Mycobacterium tuberculosis from the
complete genome sequence. Nature 393, 537–544.
based electrophoretic gel images. Genome Res 6, 1110–1117.
Dear, S., Durbin, R., Hillier, L., Marth, G., Thierry-Mieg, J. & Mott,
R. (1998). Sequence assembly with CAFTOOLS. Genome Res 8,
260–267.
Dujon, B. (1998). European Functional Analysis Network
(EUROFAN) and the functional analysis of the Saccharomyces
cerevisiae genome. Electrophoresis 19, 617–624.
Ellis, R. E., Sulston, J. E. & Coulson, A. R. (1986). The rDNA of C.
elegans : sequence and structure. Nucleic Acids Res 14, 2345–2364.
Ewing, B. & Green, P. (1998). Base-calling of automated sequencer
traces using phred. II. Error probabilities. Genome Res 8,
186–194.
Ewing, B., Hillier, L., Wendl, M. C. & Green, P. (1998). Base-calling
of automated sequencer traces using phred. I. Accuracy assessment. Genome Res 8, 175–185.
Farwell, M. A., Roberts, M. W. & Rabinowitz, J. C. (1992). The
effect of ribosomal protein S1 from Escherichia coli and Micrococcus luteus on protein synthesis in vitro by E. coli and Bacillus
subtilis. Mol Microbiol 6, 3375–3383.
Fleischmann, R. D., Adams, M. D., White, O. & 37 other authors
(1995). Whole-genome random sequencing and assembly of
Haemophilus influenzae Rd. Science 269, 496–512.
Fraser, C. M. & Fleischmann, R. D. (1997). Strategies for whole
microbial genome sequencing and analysis. Electrophoresis 18,
1207–1216.
Gardner, M. J., Tettelin, H., Carucci, D. J. & 22 other authors
(1998). Chromosome 2 sequence of the human malaria parasite
Plasmodium falciparum. Science 282, 1126–1132.
Giddings, M. C., Severin, J., Westphall, M., Wu, J. & Smith, L. M.
(1998). A software system for data analysis in automated DNA
sequencing. Genome Res 8, 644–665.
Blattner, F. R., Plunkett, G., III, Bloch, C. A. & 14 other authors
(1997). The complete genome sequence of Escherichia coli K-12.
Gillett, W., Hanks, L., Wong, G. K. S., Yu, J., Lim, R. & Olson,
M. V. (1996). Assembly of high-resolution restriction maps based
Science 277, 1453–1474.
on multiple complete digests of a redundant set of overlapping
clones. Genomics 33, 389–408.
Bodenteich, A., Chissoe, S., Wang, Y. F. & Roe, B. A. (1994).
Shotgun cloning as the strategy of choice to generate templates for
high-throughput dideoxynucleotide sequencing. In Automated
DNA Sequencing and Analysis Techniques. Edited by M. Adams,
C. Fields and J. C. Venter. San Diego, CA : Academic Press.
Bolotin, A., Sorokin, A. & Ehrlich, S. D. (1996). Mapping of the
150 kb spoIIIC–pheA region of the Bacillus subtilis chromosome
Glaser, P., Kunst, F., Arnaud, M. & 14 other authors (1993).
Bacillus subtilis genome project : cloning and sequencing of the
97 kb region from 325 degrees to 333 degrees. Mol Microbiol 10,
371–384.
Goffeau, A. & others (1997). The yeast genome directory. Nature
387 (suppl.), 5–105.
2632
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Mon, 19 Jun 2017 05:13:41
Microbial genome sequencing
Gordon, D., Abajian, C. & Green, P. (1998). Consed : a graphical
Ogasawara, N., Nakai, S. & Yoshikawa, H. (1994). Systematic
tool for sequence finishing. Genome Res 8, 195–202.
Green, P. (1997). Against a whole-genome shotgun. Genome Res
7, 410–417.
Gregory, S., Soderlund, C. & Coulson, A. (1997). Contig assembly
by fingerprinting. In Genome Mapping : A Practical Approach,
pp. 227–254. Edited by P. Dear. Oxford : Oxford University Press.
Haas, S., Vingron, M., Poustka, A. & Wiemann, S. (1998). Primer
design for large scale sequencing. Nucleic Acids Res 26,
3006–3012.
sequencing of the 180 kilobase region of the Bacillus subtilis
chromosome containing the replication origin. DNA Res 1, 1–14.
Heiner, C. R., Hunkapiller, K. L., Chen, S. M., Glass, J. I. & Chen,
E. Y. (1998). Sequencing multimegabase-template DNA with
BigDye terminator chemistry. PCR Methods Appl 8, 557–561.
Huang, X. (1996). An improved sequence assembly program.
Genomics 33, 21–31.
Isono, K. & Isono, S. (1976). Lack of ribosomal protein S1 in
Bacillus stearothermophilus. Proc Natl Acad Sci USA 73, 767–770.
Kececioglu, J. & Myers, E. (1995). Combinatorial algorithms for
DNA sequence assembly. Algorithmica 13, 7–51.
Kim, U. J., Shizuya, H., de Jong, P. J., Birren, B. & Simon, M. I.
(1992). Stable propagation of cosmid-sized human DNA inserts in
an F factor based vector. Nucleic Acids Res 20, 1083–1085.
Kim, U. J., Birren, B. W., Slepak, T., Mancino, V., Boysen, C., Kang,
H. L., Simon, M. I. & Shizuya, H. (1996). Construction and
characterization of a human bacterial artificial chromosome
library. Genomics 34, 213–218.
Klenk, H. P., Clayton, R. A., Tomb, J. F. & 48 other authors (1997).
The complete genome sequence of the hyperthermophilic,
sulphate-reducing archaeon Archaeoglobus fulgidus. Nature 390,
364–370.
Korch, C. (1987). Cross index for improving cloning selectivity by
partially filling in 5h-extensions of DNA produced by type II
restriction endonucleases. Nucleic Acids Res 15, 3199–3220.
Kunst, F., Ogasawara, N., Moszer, I. & 148 other authors (1997).
The complete genome sequence of the Gram-positive bacterium
Bacillus subtilis. Nature 390, 249–256.
Kurland, C. G. & Dong, H. (1996). Bacterial growth inhibition by
overproduction of protein. Mol Microbiol 21, 1–4.
Lander, E. S. & Waterman, M. S. (1988). Genomic mapping by
fingerprinting random clones : a mathematical analysis. Genomics
2, 231–239.
Li, P., Kupfer, K. C., Davies, C. J., Burbee, D., Evans, G. A. &
Garner, H. R. (1997). PRIMO : A primer design program that
applies base quality statistics for automated large-scale DNA
sequencing. Genomics 40, 476–485.
Luo, Y., Glisson, J. R., Jackwood, M. W., Hancock, R. E., Bains, M.,
Cheng, I. H. & Wang, C. (1997). Cloning and characterization of
the major outer membrane protein gene (ompH) of Pasteurella
multocida X-73. J Bacteriol 179, 7856–7864.
McMurray, A. A., Sulston, J. E. & Quail, M. A. (1998). Short-insert
libraries as a method of problem solving in genome sequencing.
Genome Res 8, 562–566.
Miller, M. J. & Powell, J. I. (1994). A quantitative comparison of
DNA sequence assembly programs. J Comput Biol 1, 257–269.
Nelson, K. E., Clayton, R. A., Gill, S. R. & 24 other authors (1999).
Evidence for lateral gene transfer between Archaea and Bacteria
from genome sequence of Thermotoga maritima. Nature 399,
323–329.
Offringa, R. & van der Lee, F. (1995). Isolation and characterization of plant genomic DNA sequences via (inverse) PCR
amplification. Methods Mol Biol 49, 181–195.
Petit, M. A., Dervyn, E., Rose, M., Entian, K. D., McGovern, S.,
Ehrlich, S. D. & Bruand, C. (1998). PcrA is an essential DNA
helicase of B. subtilis fulfilling functions both in repair and
rolling-circle replication. Mol Microbiol 29, 261–273.
Piper, M. B., Bankier, A. T. & Dear, P. H. (1998). A HAPPY map of
Cryptosporidium parvum. Genome Res 8, 1299–1307.
Prod’hom, G., Lagier, B., Pelicic, V., Hance, A. J., Gicquel, B. &
Guilhot, C. (1998). A reliable amplification technique for the
characterization of genomic DNA sequences flanking insertion
sequences. FEMS Microbiol Lett 158, 75–81.
Ramos-Diaz, M. A. & Ramos, J. L. (1998). Combined physical and
genetic map of Pseudomonas putida KT2440 chromosome. J
Bacteriol 180, 6352–6363.
Redenbach, M., Kieser, H. M., Denapaite, D., Eichner, A., Cullum,
J., Kinashi, H. & Hopwood, D. A. (1996). A set of ordered cosmids
and a detailed genetic and physical map for the 8 Mb Streptomyces coelicolor A3(2) chromosome. Mol Microbiol 21, 77–96.
Rose, M. & Entian, K. D. (1996). New genes in the 170 degrees
region of the Bacillus subtilis genome encode DNA gyrase
subunits, a thioredoxin, a xylanase and an amino acid transporter.
Microbiology 142, 3097–3101.
Selkov, E., Maltsev, N., Olsen, G. J., Overbeek, R. & Whitman,
W. B. (1997). A reconstruction of the metabolism of Methano-
coccus jannaschii from sequence data. Gene 197, GC11–26.
Shizuya, H., Birren, B., Kim, U. J., Mancino, V., Slepak, T., Tachiiri,
Y. & Simon, M. (1992). Cloning and stable maintenance of 300-
kilobase-pair fragments of human DNA in Escherichia coli using
an F-factor-based vector. Proc Natl Acad Sci USA 89, 8794–8797.
Siebert, P. D., Chenchik, A., Kellogg, D. E., Lukyanov, K. A. &
Lukyanov, S. A. (1995). An improved PCR method for walking in
uncloned genomic DNA. Nucleic Acids Res 23, 1087–1088.
Singh, G. B. & Krawetz, S. A. (1995).  : a software
tool for simulating contig formation for ordered shotgun sequencing. Genomics 25, 555–558.
Smith, D. R., Doucette-Stamm, L. A., Deloughery, C. & 34 other
authors (1997). Complete genome sequence of Methano-
bacterium thermoautotrophicum ∆H : functional analysis and
comparative genomics. J Bacteriol 179, 7135–7155.
Soderlund, C., Longden, I. & Mott, R. (1997). FPC : a system for
building contigs from restriction fingerprinted clones. Comput
Appl Biosci 13, 523–535.
Sorokin, A., Lapidus, A., Capuano, V., Galleron, N., Pujic, P. &
Ehrlich, S. D. (1996). A new approach using multiplex long
accurate PCR and yeast artificial chromosomes for bacterial
chromosome mapping and sequencing. Genome Res 6, 448–453.
Staden, R. (1979). A strategy of DNA sequencing employing
computer programs. Nucleic Acids Res 6, 2601–2610.
Staden, R. (1982). Automation of the computer handling of gel
reading data produced by the shotgun method of DNA
sequencing. Nucleic Acids Res 10, 4731–4751.
Sutton, G., White, O., Adams, M. & Kerlavage, A. (1995). TIGR
assembler : a new tool for assembling large shotgun sequencing
projects. Genome Sci Technol 1, 9–19.
Takeshita, S., Sato, M., Toba, M., Masahashi, W. & HashimotoGotoh, T. (1987). High-copy-number and low-copy-number
plasmid vectors for lacZ alpha-complementation and chloramphenicol- or kanamycin-resistance selection. Gene 61, 63–74.
Tao, Q. & Zhang, H.-B. (1998). Cloning and stable maintenance of
2633
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Mon, 19 Jun 2017 05:13:41
L. FRANGUEL and OTHERS
DNA fragments over 300 kb in Escherichia coli with conventional
plasmid-based vectors. Nucleic Acids Res 26, 4901–4909.
Weber, J. L. & Myers, E. W. (1997). Human whole-genome
Tomb, J. F., White, O., Kerlavage, A. R. & 39 other authors (1997).
Xu, S. Y. & Fomenkov, A. (1994). Construction of pSC101
The complete genome sequence of the gastric pathogen Helicobacter pylori. Nature 388, 539–547.
Umelo, E. & Trust, T. J. (1998). Physical map of the chromosome
of Aeromonas salmonicida and genome comparisons between
Aeromonas strains. Microbiology 144, 2141–2149.
derivatives with Camr and Tetr for selection or LacZh for
blue\white screening. Biotechniques 17, 57.
shotgun sequencing. Genome Res 7, 401–409.
Ze-Ze, L., Tenreiro, R., Brito, L., Santos, M. A. & Paveia, H. (1998).
Physical map of the genome of Oenococcus oeni PSU-1 and
localization of genetic markers. Microbiology 144, 1145–1156.
2634
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Mon, 19 Jun 2017 05:13:41